DeepSeek V4.1 Flash 1017 on one RTX 5090 builds a voxel world in 74 minutes
On a single RTX 5090 with the Maya-S quant, DeepSeek Harness built a voxel sky world from one prompt in 74.1 min, scoring 8.5/10 in browser checks at 22.9 tok/s code decode, with blocks breaking up close.
DeepSeek-V4.1-Flash-1017 on one RTX 5090 built Aetheria, a voxel sky world, in 74.1 min, on @PeasantSmith's Project Maya. It turned out interesting. His Maya-S quant, 340 GB on disk, one slot. Context 131,072 tokens. Agent: DeepSeek Harness 0.2.1-rc.2. One prompt, one try, no help. Thinking on, sampling from the DeepSeek model card: temperature 1.0, top_p 1.0. Voxel world: done in 74.1 min Tokens written: 100,575 Browser check: 8.5/10 Decode, code: 22.9 tok/s (separate speed run, thinking off) Decode, prose: 18.9 tok/s 500 words: about 35.1 s Eight requests at once: 19.7 tok/s in total What it built: floating isles with a cherry grove, a frost tower, a whale and its own camera tour between two moons. What went wrong: up close the blocks break into loose triangles, so the isles look shattered instead of solid. The browser check took 1.5 points off: the orbit-control test and half of the frame-rate test (17 fps in software GL). Open: the context was nearly full at the end (121,159 of 131,072 tokens). The last request could only write 16 tokens, because my test proxy keeps a safety margin, so the harness ended. The last saved page counts, and it loads without errors. The model card allows answers up to 262,144 tokens. The context is 131,072, so my test proxy trimmed the answer limit of 14 of the 21 requests to what still fits. Measured on Project Maya v1.1.22. Newer releases have come out since. Play the world in your browser and see the raw numbers: https://niklaslenz.com/v2/deepseek-v41-flash-5090-maya-s-mayads4/ Project Maya and the Maya-S quant by @PeasantSmith. DeepSeek-V4.1-Flash-1017 by @deepseek_ai.